Quantitative methods
taught with actual rigour
Covers the techniques that practitioners use daily - statistical modelling, supervised learning pipelines, and the uncomfortable edge cases that textbooks skip. Designed for people who already know that data rarely behaves the way the tutorial assumed it would.
Six tracks, each with a
distinct purpose
Each track addresses a specific gap - not a vague skill category. Participants work through real datasets, broken code, and decisions where the correct answer genuinely depends on context.
Statistical Inference for Analysts
Covers hypothesis testing, confidence intervals, and the assumptions that quietly invalidate results when ignored. Participants build intuition for p-values by breaking them deliberately - running underpowered tests, violating normality, and observing what actually changes in the output.
FoundationsSupervised Learning in Practice
Moves past accuracy scores into the decisions that actually matter: choosing loss functions for imbalanced targets, diagnosing leakage in time-series splits, and deciding when a simpler model is the correct one. Uses scikit-learn and real tabular datasets throughout.
Machine LearningTime Series Analysis
Addresses stationarity, seasonal decomposition, and the persistent confusion between autocorrelation and causation. Participants fit ARIMA models to messy real-world sequences and learn to interpret residuals rather than just report RMSE.
ForecastingFeature Engineering and Selection
Teaches encoding strategies for high-cardinality categoricals, interaction terms, and the difference between filter methods and embedded regularisation. The track includes a section on what to do when domain knowledge contradicts the feature importance scores.
Data PrepModel Evaluation and Calibration
Examines ROC curves, calibration plots, and the cases where a well-calibrated model is more useful than a high-AUC one. Participants learn to communicate model limitations clearly - a skill that turns out to matter more than the model itself in most production contexts.
EvaluationBayesian Reasoning for Practitioners
Introduces prior specification, posterior updating, and credible intervals without requiring a statistics PhD as a prerequisite. Covers PyMC for applied modelling and spends considerable time on prior sensitivity - because the choice of prior is rarely as neutral as it appears.
AdvancedQuestions about which
track fits your
situation
Each track has defined prerequisites listed on the learning programme page. If the description sounds right but the prerequisites look uncertain, reach out - the honest answer is usually more useful than a sales conversation.